Stability and Similarity Detection for the Biologically Inspired Temporal Pooler Algorithms

Abstract

The biology of the brain can inspire the development of a wide range of intelligent information processing algorithms. Spatiotemporal processing allows the brain to aggregate and represent information on multiple time and space scales. This paper examines temporal pooling, a biologically inspired spatiotemporal processing algorithm, in the task of creating stable sequential representations of objects while preserving their relative similarity. We compare two temporal pooling algorithm implementations: an existing Hierarchical Temporal Memory framework implementation called Union Temporal Pooler and our proposed Sandwich Temporal Pooler. We evaluate the temporal poolers on synthetic sequences using various aggregating metrics. It is shown that both implementations preserve high similarity; however, Union Temporal Pooler has better ability to preserve low and moderate similarity, while the proposed Sandwich Temporal Pooler features better output representation stability and computational efficiency.